Using constraint-based metabolic modeling to elucidate drug-induced metabolic changes in a cancer cell line

Since the discovery of the Warburg effect, metabolism has become a crucial cancer hallmark, for cancer cells change the activity of metabolic pathways to satisfy the demand of biomolecules and energy required to sustain continuous growth. Specially, gastric cancer (GC) still remains a growing burden of society, and often require the development of targeted and personalized interventions, including the use of combinatorial drug therapies that allow for synergistic interactions. However, the mechanism underlying synergistic interactions between multiple drugs remain poorly understood. The objective of this study is to use genome-scale metabolic models (GEMs) and transcriptomic data to investigate potential synergistic mechanisms in the GC cell line AGS following treatment with three kinase inhibitors and their combinations. To explore the metabolic phenotype of AGS cells, we curated and used a set of descriptions of metabolic functions (metabolic tasks) based on the latest iteration of the Human-GEM, and we used the Task Inferred from Differential Expression (TIDE) framework to infer the metabolic changes in AGS cells. Additionally, we expanded and complemented the TIDE framework by integrating essential genes from metabolic tasks, allowing for a more robust analysis. The results revealed a complex metabolic response in AGS cells after kinase inhibitor treatments, particularly in amino acid and nucleotide metabolism, and characterized by the down-regulation of metabolic functions. Using two definitions of synergistic effects, we identified significant metabolic shifts in combinatorial treatments, including a combination-specific increase in ornithine levels, potentially linked to an antiproliferative effect on AGS cell growth. These findings highlight the potential of using transcriptomic data to infer the metabolic phenotype of AGS cells and the role of metabolic alterations in cancer progression. In addition, we developed a Python library for Metabolic Task Enrichment Analysis, making the TIDE frameworks accessible to other researchers.